JPH0536352B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a voice recognition control method for controlling a crane using voice recognition.

(Prior Art) Generally, in a crane control device that controls a crane using a voice recognition device, voice commands have various functions, and depending on the type of voice command, execution is performed only when it is reliably recognized. This includes commands that must be executed immediately even if there is some doubt as to whether recognition has been performed correctly. For example, the voice command word "stop" to instruct a crane to stop while it is moving.
If you input , even if the registered standard pattern and the input voice pattern are not very similar, and even if the input voice may not be "stop", it will be regarded as "stop" and the crane will stop. The action must be performed immediately. This is because if a stop operation is not performed when a stop is necessary, the crane may collide with another object and cause damage, the crane itself may break, or people may be injured. On the other hand, in the case of the voice command word ``Rifmag release'' that instructs the work of releasing a steel load that has been lifted and transferred by the crane's electromagnet and releasing it from the electromagnet attraction, the input voice must be ``Rifmag release''. Only when this is certain is it necessary to treat this as a ``rifmag release'' and perform the task of deactivating the crane's electromagnet to release the cargo from suction. This is because if the suction of the electromagnet is accidentally released when the load is in a position where it should not be unloaded, the load may fall and cause a serious accident.

Conventionally, an input speech pattern is compared with various registered standard patterns, and the standard pattern with the smallest difference (called a score value) between the two and whose difference is less than a certain threshold is selected. However, as mentioned above, depending on the type of voice command, more gradual or more reliable recognition is required, but no such distinction was made.

(Problems to be Solved by the Invention) The present invention solves the above-mentioned problems of the prior art, and provides a crane that presets the recognition weight according to the type of command and performs safer control. The present invention provides a voice recognition control method in a control device.

(Means for solving the problem) In order to achieve the above purpose, M multiple registrations are made in advance for each command word as a standard pattern, and when the command word is input by voice, the input voice pattern N items (N≦
A command word corresponding to the standard pattern M) is adopted as a recognition command word, and when all of the N recognition command words match the input command word, an operation is performed based on the command, and N recognition command words. When at least one of these matches an input command word, an operation is performed based on that command, and the other is set for each command word.

(Function) According to the above configuration, for a type of command that would cause serious consequences if executed while misrecognizing the command, such as "Rifmag release", the N commands extracted in order from the one with the lowest score value are When all the recognized command words match the input command words (AND is established), more accurate recognition is performed so that the command is operated based on that command, and when the command is missed and execution is neglected, such as "stop", For commands that have serious consequences, when at least one of the N recognized command words matches the input command word (OR holds), the system operates based on that command so that the command is not overlooked. do. By setting each command word in advance in this way, safer operation of the crane apparatus is possible.

(Example) Hereinafter, an example will be described in detail with reference to the drawings. First, an outline of a crane control system to which the voice recognition control method of the present invention is applied will be explained with reference to FIG. The system includes a recognition device control panel 1, a crane control panel 2, a crane 3, a wireless device mobile station 4, a headset microphone 5, and a wireless modem 6.
and a console 7. The recognition device control panel 1 includes a wireless device fixed station 11, a voice recognition device 12, a voice response device 13, an interface relay panel 14, and a wireless modem 15.

The wireless device fixed station 11 is a device that receives a signal sent from the wireless device mobile station 4 and demodulates it into a voice command signal, or modulates and transmits a voice signal from the voice response device 13.

The voice recognition device 12 works together with the voice response device 13 to interpret analog voice information from the wireless device fixed station 11, compare it with pre-registered voice command information (standard pattern), and find a match. The content of the voice command is identified based on the level of the voice command, and the corresponding operation command word for controlling the crane is output.

The voice response device 13 generates a voice signal in response to the voice recognition result in order to confirm whether the voice recognition has been performed correctly.
This audio signal is transmitted to the driver via the wireless device fixed station 11 and the wireless device mobile station 4.

The interface relay board 14 converts the output from the voice recognition device 12 and drives a relay.

The wireless modem 15 performs transmission and reception with a wireless modem 6 connected to a console 7 on the ground, demodulates the received command signal from the console 7 and transmits it to the command decoding section of the voice recognition device 12, or performs voice recognition. It modulates the signal from the device 12 in order to transmit it wirelessly to the console 7.

The crane control panel 2 has a contactor that is controlled by the relay output of the interface relay panel 14 and the detection signal from the limit switch 33 of the crane 3, and drives the motor 31, brake 32, etc. for the crane 3. It also turns on and off power to alarms, illumination lights, etc. (not shown).

The wireless mobile station 4 is carried by the driver, modulates voice commands from the headset microphone 5, transmits the modulated voice commands to the recognition device control panel 1 wirelessly, and also transmits the modulated voice commands to the recognition device control panel 1.
This is a wireless device that receives the voice output of the voice response device 13 transmitted through the voice response device 1 and transmits it to the driver through the earphone of the headset.

The console 7 and wireless modem 6 are installed on the ground to enable console input in addition to voice input using the headset microphone 5 and wireless device mobile station 4 described above. Used for temporary stops and other operations.

Next, the basic operation of the crane system configured as described above will be explained.

The voice command received by the wireless device fixed station 11 is recognized by the voice recognition device 12, and
In response to the recognized command, various control signals necessary to control the operation of the crane are output. Its functions are realized by the operation of a crane control program executed by a microcomputer for the voice recognition device. FIG. 3 shows an outline of the flow of the operation.

After starting the program, input necessary constants such as inching constant COn, steady rest constant DOn, etc. from the console 7 as initial value settings for the program so that the crane equipment used will respond appropriately to voice commands. do.

Next, it starts the voice recognition program and waits to receive commands from the driver.

First, when the operator gives a voice command to turn on the crane, the voice recognition device 12 recognizes the command and sends a corresponding command signal to the interface relay board 14, which activates the power relay. Turn on the power to the crane control panel 2 and crane 3, make the crane operational,
Wait to receive the next command from the driver.

When the voice command from the driver is recognized, the result is stored in the response buffer RESBUF.
It determines what the stored command is, moves to the corresponding operation subroutine, and executes it. The present invention is characterized by the determination of this voice command, which will be described in detail later.

Recognized voice commands include "power on", "power off", "alarm", "lighting on", "lighting off", etc., which can immediately and unconditionally execute the operation of the crane equipment. For those that do, the process immediately moves to the operation subroutine and is executed. In other words, the response buffer
When the RESBUF content is "power on", it is as described above, and when it is "power off", all outputs of the voice recognition device are set to zero, the crane equipment is powered off, and the subroutine that returns to the beginning is executed. . When there is an "alarm", "lighting on", or "lighting off", the system moves to the alarm subroutine, lighting on subroutine, and lighting off subroutine, respectively, which are executed and output signals to control the alarm, power on/off, etc. Then, the device returns to the standby state to receive the next voice command.

On the other hand, commands that indicate the direction of operation such as "eastward", "westward", "southward", "northward", "upward (winding up)", "downward (downward)", etc. , in the case of commands that instruct the crane to operate in combination with commands such as "start,""stop,""jog," and "steady rest," which instruct the execution of movements in the indicated direction. , the command instructing the movement direction is temporarily stored in the movement direction instruction buffer BUF, and returns to the standby state to receive the next voice command.

If the execution command is "start", "stop", "jog", "steady rest", etc., which is used in combination with the motion direction instruction command, the start subroutine, crane stop subroutine, and inching motion subroutine are used. are processed respectively. Note that the start subroutine is a routine that generates an output for causing the crane to start an operation according to the contents of the operation direction instruction buffer BUF, and returns to the state of waiting for the next voice command. The crane stop subroutine is for stopping a crane that is operating in accordance with the contents of the movement direction indication buffer BUF, and clearing the contents of the movement direction indication buffer BUF. The inching operation subroutine is for performing inching or steady rest operation.

In the present invention, first, M multiple registrations are made for each command word as registered standard patterns stored in the storage section of the speech recognition device 12. For example, if command word a is registered M times and the group is A, then A=(a ₁ , a ₂ , a ₃ , . . . , a _M ).

Next, when a command word is input by voice, the input voice pattern is compared point by point with various registered standard patterns, and N standard patterns are matched in order from the one with the smallest score value, which is the difference between each. The command word to be recognized is extracted as the recognized command word. Here, it is assumed that N≦M. Now, if the command word a is generated and input by voice, and adopted in the order of score values up to the Nth recognition, the following combinations of N+1 sets f are possible.

0 f ₀ = (N pieces of a) 1 f ₁ = (N-1 pieces of a, 1 piece of b) 2 f ₂ = (N-2 pieces of a, 2 pieces of b) − − − − − − N-1 f _N-1 = (1 a, N-1 b) N f _N = (0 a, N b) Here, b is a pronunciation that is relatively similar to a. The command word is Further, there may be a case where the score value of the extracted b is smaller than that of a. Furthermore, a command word other than b may be extracted.

Therefore, only when all of the adopted N recognition command words match the input command word, that is, when the AND from the first recognition to the Nth recognition is established, as shown in FIG. When at least one of the N recognized command words matches the input command word, that is, as shown in Figure 1b, the OR from the first recognition to the Nth recognition holds. When a command is issued, an action is performed based on the command, and each command word is set individually. As shown in the above set of f, as shown in Fig. 4, AND is true only when f ₀ ,
All other cases from f ₁ to f _N are rejected. This is a case that requires more accurate recognition and is applied to command words such as ``Rifmag release''. On the other hand, OR is established in the case of f ₀ to f _N-1 , and the input command word is assumed to be recognized and executed immediately, and is rejected only in the case of f _N. This applies to commands such as "stop" when the command should not be overlooked rather than accurate recognition.

As mentioned above, by setting M and N appropriately, AND,
By operating the OR programmatically, safety can be dramatically increased.

(Effects of the Invention) As explained above, according to the present invention, when controlling a crane using voice recognition, commands of the type that will cause serious consequences if executed without being recognized incorrectly, such as "Rifmag release", can be used. performs more accurate recognition so that the operation based on the command is performed only when all the recognition command words from the first recognition to the Nth recognition match the input command word (AND is established), and the "stop" command is executed. For commands of a type that would lead to serious consequences if the command is overlooked and neglected, at least one of the recognized command words from the first recognition to the Nth recognition matches the input command word. When (OR is established) the operation is performed based on the command word and the command is not overlooked, so the safety of the operation of the crane device can be significantly improved.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the AND and OR operation flow of an embodiment of the present invention, FIG. 2 is a schematic block diagram of a crane control system to which the voice recognition control method of the present invention is applied, and FIG. FIG. 2 is a diagram showing the basic operation flow of the system, and FIG. 4 is a diagram showing the recognition range and reject range of AND and OR in the Nth recognition. 1... Recognition device control panel, 2... Crane control panel, 3... Crane, 4... Wireless device mobile station, 5
...Headset microphone, 6...Wireless modem, 7
... Console, 11 ... Wireless device fixed station, 12
...Voice recognition device, 13...Voice response device, 14
...Interface relay board, 15...Wireless modem.

Claims (1)

[Claims] 1. In a voice recognition control method for a crane control device that identifies voice commands and drives a crane device according to the identification results, M multiple registrations are made in advance for each command word as a standard pattern, and the command word When a voice input is performed, N pieces (N≦
A command word corresponding to the standard pattern M) is adopted as a recognition command word, and when all of the N recognition command words match the input command word, an operation is performed based on the command, and N recognition command words. A voice recognition control method for a crane control device, characterized in that when at least one of the input command words matches an input command word, an operation is performed based on the command, and a setting is made for each command word.